This branch tries to (and ultimately fails) speed up the review input…

…s metadata endpoint in Hintr. Profiling the code revealed these Naomi functions were the slow parts. After optimising these they are many times faster on the test data and the MWI data however where we ultimately fail is DRC for which there is only a slight speed up because the majority of the time for it is spent calculating the other plot types like art_new_total, art_new_adult_m, etc. We have a solution for this that will probably get done in a PR after the plot refactor. For now, the main changes in this PR:

* drop_geometry arg to aggregate functions to skip unnecessary work (these functions were trying to drop geometry after it had already been dropped so wasting a lot of time reading the files)
* small test fix, we should be checking `check3` (defined in the lines just above the diff) and not `check1`
* new algorithm for aggregating up data, it is just a for loop that works level by level to aggregate up the data
* same new algorithm is also applied to calculating `missing_map`

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: naomi
 Title: Naomi Model for Subnational HIV Estimates
-Version: 2.9.27
+Version: 2.9.28
 Authors@R:
     person(given = "Jeff",
            family = "Eaton",